Disclaimer: This report has been written for the authors learning purposes only.

Brief

To inform the planning and provision of cancer treatment services, analyse breast cancer incidence data from the NHS Borders health board. Findings should be presented in a report of 1-2 pages, highlighting any significant insights and identified trends.

Introduction

Between 1997-2021, incidences of breast cancer in males made up less than 1% of total breast cancer incidences. This report will therefor focus on instances of breast cancer among females.

cancer_incidence_borders %>%
  filter(cancer_site != "All Cancer Types",
         sex == "All") %>%
  group_by(cancer_site) %>%
  summarise(total_incidences = sum(incidences_all_ages)) %>%
  arrange(desc(total_incidences)) %>%
  head(3) %>% 
  gt() %>%
  tab_header(title = "Total Cancer Incidences") %>% 
  cols_label(
    cancer_site = "Cancer Site",
    total_incidences = "Total Incidences") %>%
  tab_options(column_labels.font.weight = "bold",
              table.align = "left")
Total Cancer Incidences
Cancer Site Total Incidences
Non-Melanoma Skin Cancer 6174
Basal Cell Carcinoma Of The Skin 4049
Breast 2614
 
cancer_incidence_borders %>%
  filter(cancer_site == "Breast",
         sex != "All") %>%
  group_by(sex) %>%
  summarise(total_incidences = sum(incidences_all_ages)) %>%
  arrange(desc(total_incidences)) %>%
  head(3) %>% 
  gt() %>%
  tab_header(title = "Breast Cancer Incidences by Sex") %>% 
  cols_label(
    sex = "Sex",
    total_incidences = "Total Incidences") %>%
  tab_options(column_labels.font.weight = 'bold',
              table.align = "left")
Breast Cancer Incidences by Sex
Sex Total Incidences
Females 2598
Male 16

According to NHS Borders data, breast cancer among females has the highest number of incidences, highest mean crude rate and highest mean European age-standardised rate (EASR) of any cancer type. See below for a more detail on how the EASR is calculated.

cancer_incidence_borders %>%
  filter(cancer_site != "All Cancer Types",
         sex == "Females") %>%
  group_by(cancer_site) %>%
  summarise(total_incidences = sum(incidences_all_ages)) %>%
  arrange(desc(total_incidences)) %>%
  head(3) %>% 
  gt() %>%
  cols_label(
    cancer_site = "Cancer Site",
    total_incidences = "Total Incidences") %>%
  tab_options(column_labels.font.weight = 'bold')
Cancer Site Total Incidences
Breast 2598
Non-Melanoma Skin Cancer 2519
Basal Cell Carcinoma Of The Skin 1882
 
cancer_incidence_borders %>%
  filter(cancer_site != "All Cancer Types",
         sex == "Females") %>%
  group_by(cancer_site) %>%
  summarise(mean_easr = mean(easr)) %>%
  arrange(desc(mean_easr)) %>%
  head(3) %>%
  gt() %>%
  cols_label(
    cancer_site = "Cancer Site",
    mean_easr = "Mean EASR") %>%
  tab_options(column_labels.font.weight = 'bold')
Cancer Site Mean EASR
Breast 161.3640
Non-Melanoma Skin Cancer 150.3996
Basal Cell Carcinoma Of The Skin 113.9178

Health Board Comparison

To understand how these rates compare to other health boards in Scotland, we can visualise a five year summary of the EASR - which is a “weighted mean of the age-specific rates where the weights are taken from the population distribution of a standard population; the ASR is expressed per 100,000.” (European Commission, 2023)

geo_summary %>% 
  ggplot(aes(fill = easr)) + 
  geom_sf(colour = "white", linewidth = 0.04) +
  labs(
    title = "Female Breast Cancer EASR (2017-2021)",
    subtitle = "By NHS Health Board",
    fill = "EASR") +
  scale_fill_distiller(palette = "Blues", direction = +1) +
  theme(panel.background = element_rect(fill = "white"),
        axis.text.x = element_blank(),
        axis.text.y = element_blank(),
        axis.ticks = element_blank(),
        rect = element_blank(),
        axis.title.x = element_blank(),
        axis.title.y = element_blank())

NB - Unfortunately data for the individual health boards NHS Western Isles, NHS Shetland and NHS Orkney was not available at the time of report completion.

five_year_summary %>%
  select(hb, cancer_site, sex, year, easr) %>%
  filter(sex == "Females",
         cancer_site == "Breast",
         hb != "GR0800001") %>%
  left_join(geography_codes, "hb") %>% 
  select(hb_name, year, sex, cancer_site, easr) %>% 
  arrange(desc(easr)) %>% 
  gt() %>%
  cols_label(
    hb_name = "NHS Health Board",
    year = "Year(s)",
    sex = "Sex",
    cancer_site = "Cancer Site",
    easr = "EASR") %>% 
  tab_options(column_labels.font.weight = 'bold') %>%
  data_color(columns = easr, palette = "Blues")
NHS Health Board Year(s) Sex Cancer Site EASR
NHS Dumfries and Galloway 2017-2021 Females Breast 174.6153
NHS Lothian 2017-2021 Females Breast 172.3179
NHS Forth Valley 2017-2021 Females Breast 171.6585
NHS Lanarkshire 2017-2021 Females Breast 169.1486
NHS Greater Glasgow and Clyde 2017-2021 Females Breast 168.8007
NHS Borders 2017-2021 Females Breast 164.8136
NHS Fife 2017-2021 Females Breast 164.4207
NHS Tayside 2017-2021 Females Breast 163.4222
NHS Highland 2017-2021 Females Breast 162.5039
NHS Ayrshire and Arran 2017-2021 Females Breast 157.0019
NHS Grampian 2017-2021 Females Breast 156.2987

Hypothesis Test

Question: Is the greater number of female breast cancer incidences in “peak years” (1999, 2002, 2005, 2008, 2011, 2014, 2017) compared to “non-peak years” (1997, 1998, 2000, 2001, 2003, 2006, 2007, 2009, 2010, 2012, 2013, 2015, 2016, 2018, 2019) statistically significant?

cancer_incidence_borders_sample <- cancer_incidence_borders %>%
  filter(sex == "Females", cancer_site == "Breast") %>%
  select(id, cancer_site, sex, year, incidences_all_ages) %>% 
    mutate(peak = case_when(
    year == 1999 ~ "peak",
    year == 2002 ~ "peak",
    year == 2005 ~ "peak",
    year == 2008 ~ "peak",
    year == 2011 ~ "peak",
    year == 2014 ~ "peak",
    year == 2017 ~ "peak",
    TRUE          ~ "standard"
    )
  )

observed_stat <- cancer_incidence_borders_sample %>% 
  specify(incidences_all_ages ~ peak) %>%
  calculate(stat = "diff in means", order = c("peak", "standard"))

null_distribution <- cancer_incidence_borders_sample %>% 
  specify(response = incidences_all_ages, explanatory = peak) %>%
  hypothesize(null = "independence") %>%
  generate(reps = 1000, type = "permute") %>% 
  calculate(stat = "diff in means", order = c("peak", "standard"))

p_value <- null_distribution %>%
  get_p_value(obs_stat = observed_stat, direction = "right")

Test Used: Two Sample Mean Test (Independent)
Significance Level: 0.05

H0: \(\mu{1}\) - \(\mu{2}\) = 0
H1: \(\mu{1}\) - \(\mu{2}\) > 0

Result: Based on a bootstrapped NULL distribution, a very low p-value which is less than 0.05 is returned. We therefor reject H0 in favor of H1 with evidence suggesting that there is a statistically significant increase in the mean number of female breast cancer incidences in “peak years”.

Incidences by Age

[fig. 2]

fig2_plot <- five_year_summary_long %>%
  filter(cancer_site == "Breast",
         sex == "Females") %>%
  ggplot() +
  geom_col(aes(x = age, y = incidences, 
               text = paste0("<b>Age:</b> ", age, "<br>", "<b>Incidences:</b> ", incidences, "<br>")),
           fill = "#0391BF") +
  theme(axis.text.x = element_text(angle = 45, vjust = 0.5)) +
  labs(
    x = "\n Age",
    y = "Incidences\n",
    title = "Total Female Breast Cancer Incidences by Age") +
  theme(panel.background = element_rect(fill = "white"),
        panel.grid = element_line(colour = "grey90"))

ggplotly(fig2_plot, tooltip = "text") %>%
  layout(title = list(text = paste0("<b>Total Female Breast Cancer Incidences by Age</b>",
                                    "<br>",
                                    "<sup>",
                                    "NHS Borders: 1997-2021",
                                    "</sup>")))

What does this visualisation tell us?

  • The majority of breast cancer incidences in females appear to be between those aged between 50 and 79.

Why might these age groups see increased incidence numbers?

  • Currently only women between the ages of 50 and 70 are routinely screened (NHS National Services Scotland, 2022).

NHS Borders Population Projections:

Females 50+ 2021: 29889

Females 50+ 2041: 31148 (4.21225% increase)

(National Records of Scotland, 2023)

Conclusions / Recommendations

  • Screening data should be reviewed to establish if the resulting back-log from COVID-19 has been cleared in order to establish whether a further increase in incidences should be anticipated in 2022.

  • Resources should be allocated according to the observed trend of increased incidences every three years

  • Research/Analysis should be conducted to further understand and confirm any reason for this trend, including any links to screening schedules.

  • Research/Analysis should be conducted to establish whether increased incidence with age is in any way the result of current screening criteria and if therefor screening criteria should be widened.

  • Long term service planning should take into consideration the ~4% projected population increase of the female 50-70 demographic in NHS Borders, as projected by the National Records of Scotland.

Data Sources